DEV Community: Saloni Agarwal

NanoGem Nail Art Studio 💅🏻

Saloni Agarwal — Sun, 14 Sep 2025 16:37:13 +0000

This is a submission for the Google AI Studio Multimodal Challenge

What I Built

Introducing NanoGem Studio, a delightful and engaging nail art experience built with the power of NanoBanana and Google AI Studio.
This applet allows you to virtually try on different nail art designs from a palette, try custom studio views that further enhances the beauty of nails, and also create your own nail art look using natural language prompts.It also offers a gallery section to save your favorite designs for future use.
So, the next time when you sit with your nail art kit at home or pay a visit to some studio, you will have your own personalized collection of nail art designs to get that perfect dreamy look on your nails.

Demo

Applet link: live app

Applet demo

Applet Screenshots

How I Used Google AI Studio

Ideation and prompting: Initially, I wrote a simple prompt that let me layout my idea and build the core MVP easily, with no errors.
Iterative feature addition and testing: As I continued building the app, I kept prompting the code assistant with tiny features and UI fixes. The Auto-fix feature is something that I personally loved a lot, it showed up whenever there was some error in running the code.
Custom prompts for gemini 2.5 flash-image-preview: I wrote many prompts for designing and testing the palette and custom studio views which were passed on to the image generation and editing model API.
Code remix: Certain UI elements of the applet which were generated by the code assistant did not look very good, so I wrote some code here and there in different component files to get the desired UI.
Applet Deployment: Using the Cloud Run feature, I deployed the app directly from AI Studio. In the Google Cloud console website, I set the service scaling to auto in the Cloud Run, services tab to ensure that the applet runs smoothly.

Multimodal Features
Text-to-Image: For the Palette screen in the applet, I designed 5 descriptive text prompts to define a variety of nail art designs. I leveraged the gemini 2.5 flash-image-preview API to get the best generated images from these prompts. The Create Your Own design feature in the applet also uses this API to generate custom user-defined nail art designs.
Image + Text-to-Image(Editing): The following custom studio views utilize this multimodal feature provided by gemini 2.5 flash-image-preview API:
- Sun-kissed view: On the generated nail art design, I passed in a text prompt to add a sun-kissed look to the image which lets users see how a certain design glows in a bright sunlit area.
- Accessories add-on view: This view option adds a bracelet on the wrist of user which coordinates with the generated nail art design, giving a more complete and stylized look.

Conclusion

I had an incredible amount of fun into crafting this experience, because this project gave me the opportunity to make something that I personally care a lot about. NanoGem Studio is not just an experience , it plays the role of your own Nail Art Style Assistant that delivers the best designs at your fingertips.
Let me know in the comments which palette look you loved the most, and what other features you liked!
Cheers🥂 💅🏻

Meet Pulse - the ultimate skill coach agent ✨

Saloni Agarwal — Mon, 07 Jul 2025 06:10:00 +0000

This is a submission for the Runner H "AI Agent Prompting" Challenge

What I Built

Say hello to Pulse, your smart AI skill coaching agent that allows you to learn any skill with a robust and relentless action plan in just 2 weeks.

Key features

Integrates Surfer H, Google Doc and Google Calendar connections
Generates detailed personalized learning plans
Automated task reminders for 2 weeks aligned with user's learning pace

Demo

See the Pulse Runner H agent in action here

Screenshots 📺

How I Used Runner H

The prompt ✨

**You are Pulse, an AI coaching agent that helps users  learn any {{skill}} in 2 weeks using a detailed action feedback loop {{plan}} through  productive, smart and efficient ways.**

You have access to the following tools (with the user’s permission):
* Google Docs
* Google Calendar
* Surfer H agent


# TASKS

## A. Follow the below steps in workflow to design the {{plan}}:

### 1. Ask the user to answer the following before moving to step 2:
 1.  what {{skill}} you want to learn?
 2. Why do you want to learn the {{skill}}?
 3. What is your learning pace ?
 4. Any past experience or basic knowledge of {{skill}} ?
 5. How many hours can you put in daily?

### 2. Design a detailed  2 week {{action}} feedback loop {{plan}} based on the instructions below: 
a. Use the Pareto principle: focus on the 20% of concepts that give 80% of the results.
b. Assign daily tasks to the user.
c. Search the web for the best publicly available learning resources to master the {{skill}}, make sure these resources are in line with the learning pace of user. 
d. Now, draft the plan in a structured and neat format.

### 3. Add the {{plan}} to the Google Docs of user. Populate the doc with the following structure:
* ## Document Heading: The Two-Week Plan 🚀 - crafted by Pulse, your AI skill coach
* Objective: Master {{skill}} proficiently.
* Daily Mission Control:
      (Use the format below until **Day 14**):
      * Day 1:
      -Task 1
      -Task 2 
      - Feedback
* Useful Tips:
    * Provide  3-4 strategies to maintain focus and productivity throughout the learning process.

### 4. Display the message in chat,” Google Doc designed successfully”.

—-Exit workflow A—-

—-Enter workflow B—-

## B.Follow the below steps in workflow:

### 1. Integrate every {{action}} in the Google doc as a reminder for 14 days in the Google Calendar of  user according to their {{learning pace}}.
### 2.  Ask the user for confirmation of schedule in Google calendar from the user and allow them to make changes if any, 
then proceed to step 3.
### 3. Display the Google Calendar schedule.
### 3. Suggest some further actions aligned with learning and implementing the {{skill}} in a smooth and concrete manner.

—-Exit workflow B—-

—-Enter workflow C—-

## C. Follow the below steps in workflow: 

### 1. Display the final learning {{plan}}
### 2. Use an encouraging and positive tone to write an ending  {{note}} to user that they can successfully learn the {{skill}} in 2 weeks without having to feel stressed or pressured  and having the maximum fun possible. 
### 3. End the {{note}} with a powerful quote.
### 4. Write the {{note}} in markdown italics format.
### 5. Display the note.

—-Exit workflow C—-

# END TASK
Upon the successful execution of all workflows, prompt the user in chat with the final message:
“I have designed the 2 week plan for learning {{skill}} as you asked, if you wish to make any changes, I would love to help, sincerely Pulse —your smart AI skill coach 😊”

——ALL TASKS COMPLETED SUCCESSFULLY——


# RULES 
1. The plan must have tasks outlined for **14 days**.
2. After completing each workflow display the message in chat, “Workflow executed successfully”. Eg: Workflow A executed successfully.
3. The Google doc must be readable and organized as instructed.

Key Steps in the agent workflow

Learner profile: The user is asked to specify the skill and its target goal, their basic knowledge of skill, learning pace and daily commitment towards the plan.
Web Search: Surfer H agent searches the web for the best learning resources according to the specified skill
Detailed Plan design: the 2 week learning plan is stored in a Google Doc which the learners can refer to regularly.
Automated task reminders: the Google Calendar connection is used to set up daily reminders for the tasks mentioned in the doc.
Final ending note: the agent crafts a positive and motivating message for the users so that they can successfully navigate through their learning journey.

Use Case & Impact

Real World Applications

1. Creators and Indie Builders
Use-case - Content creators/indie hackers learning new tools(e.g. SEO, video editing)

Keeps them focused through structured daily task nudges in fast-paced environments.
Helps them up skill real fast in different domains.

2. College and High School Students
Use-case - Students who want to build portfolios, prepare for internships, or explore side skills (e.g. Canva, coding, writing)

Encourages daily goal based learning
Reduces stress and burnout since learning plans are realistic and goal-based

3. Corporate Upskilling and Internal Training
Use-case - Companies can use Pulse to train employees on new tools/processes quickly

Offers accelerated learning for fast onboarding or role transitions
Integrates seamlessly into work calendars

How it improves existing processes

✅ 1. Replaces Passive Learning with Action-Based Loops
✅ 2. Automates Structure Where None Exists
✅ 3. Improves Retention and Completion Rates
✅ 4. Crafted meticulously for modern fast-paced environments and ambitious learners

Social Love 💖

If you like this agent, show some love!!!
// Detect dark theme var iframe = document.getElementById('tweet-1942102528453476527-702'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1942102528453476527&theme=dark" }

Why I built this

I believe in modern competitive and fast-paced environments many people struggle to learn new things quickly and are often crushed under the weight of feeling behind as they see others speed run through new skills and tools and do it all effortlessly. And, yes I struggle with this too.
So this led to the idea of making a tool that can break a skill into time bound micro actions spread over a span of 2 weeks coupled with automated planning.

Pulse is dedicated to all the ambitious and curious learners around the world who want to dazzle their way into mastering anything they want.

A sincere thanks to Runner H and dev.to for this wonderful opportunity!

Compiling 2025: learn, build, grow...

Saloni Agarwal — Sun, 02 Feb 2025 19:40:33 +0000

This is a submission for the 2025 New Year Writing challenge: Compiling 2025.

Introduction

You might have often read this line

You overestimate what you can do in a day, but underestimate what you can do in a year

Well, I have always underestimated what I could accomplish in a year, which is why most of my goals didn't take off last year.It left me extremely disheartened. After a lot of self-reflection and thinking, I realized that I need to place a bet on all the things that matter to me and take action on them relentlessly to produce the best possible outcome. No matter how difficult the road is,I need to keep going, keep running and make things HAPPEN!

On that note, lets take a look on my 2025 roadmap...

Main goals 🎯

Internship at an AI startup
This is really important for kickstarting my career in AI.
Key skills to master-
PyTorch, deep learning, neural networks
ML algorithms, scaling, optimization
MLOps, CUDA, kernels. And lastly, making sure my projects highlight these skills in the best way possible.
Develop strong skills in NLP
I started learning NLP in mid November 2024. Honestly, the subject seemed very dry at first but gradually it started making sense to me. The learning curve dropped in December, but now it's coming back up again. And so I am going to spend more hours into it in the next three months henceforth.
Do meaningful open source contributions
I suck at open source and I mean it. But it's high time to do something about it. This is why I joined Quira last year(sneaking in a little shoutout to the community 🤭), where one can explore the true beauty of open source by participating in bi-weekly hackathons, making PRs in open source orgs and even earn bounties!
Okay coming back to the point, I need to make more open source contributions this year and write code that can help other developers .
Become a cool developer - one that can always get things done
Achieving the above 3 goals leads to this ultimate one goal.

A small list of resources for folks interested in AI -

LLMs

rasbt / LLMs-from-scratch

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

Build a Large Language Model (From Scratch)

This repository contains the code for developing, pretraining, and finetuning a GPT-like LLM and is the official code repository for the book Build a Large Language Model (From Scratch).

In Build a Large Language Model (From Scratch), you'll learn and understand how large language models (LLMs) work from the inside out by coding them from the ground up, step by step. In this book, I'll guide you through creating your own LLM, explaining each stage with clear text, diagrams, and examples.

The method described in this book for training and developing your own small-but-functional model for educational purposes mirrors the approach used in creating large-scale foundational models such as those behind ChatGPT. In addition, this book includes code for loading the weights of larger pretrained models for finetuning.

Link to the official source code repository
Link to the book at Manning…

View on GitHub

PyTorch
CUDA

Side Projects ❤️‍🔥

Ah, side projects. They are my favorite to be honest.Whenever I undertake a new side project, it's like stepping into the unknown. Now,I started learning SwiftUI in July 2024 and made 4 apps so far. It was a wonderful journey one that has led me to making an app for the Swift Student Challenge this year.This is my most important side project so far, and I hope to deliver a really good app at the end whether I win or not. But hey, thats not all. I plan to make at least 5 projects in Robotics this year. With the recent advancements in AI, it is evident that robotics is going to gain momentum in the next decade hence, I definitely want to be a part of the loop. Self-teaching myself hardware is going to be hard, but I don't want to miss out on the fun of being a MAKER. In fact, the first project is going to be powered by a Raspberry Pi Pico.

Here are some cool Youtube channels that you can check out:

Estefannie ,HannahMakes - for robotics and IoT.
twostraws, Apple developer - for iOS dev

Skills to learn 🖋️ 🎹

Writing
As I type away this blog, my writing skills are definitely getting a major upgrade 🤭. Had no plans for this honestly but it's worth a shot. So I am gonna set a goal of writing at least 5 blogs this year and in the process find out if writing is my cup of tea or not.
Playing a musical instrument
Being a big fan of soundtracks and instrumental music, this is something that I have to make time for. I still don't know what music instrument it's going to be but I guess we'll figure it out.

New habits 🍀

I feel this is something that I have often ignored. As time passes, we tend to develop habits that seem quite harmless on the surface but can lead to huge troubles in the future. Personally, I feel a habit at its root is just a way of thinking. You need to be dead serious about building good habits and letting go of bad ones. So here's a list of habits that I wanna put to production in 2025 💪🏻:

Getting up at 5:30 am in mornings
Morning run for 20 mins
Reading everyday at night
Praying in the morning
Journalling each day of my life

Conclusion

Alright, if you are still reading my blog, you might say, "Oh gosh, thats a really long list of things!"
Right it is, but its the only way to I wish to spend my each day this year. I might fail while tackling the above roadmap but its not going to me down. For me, 2025 is not just about growing as a developer but as a person too.

So, for all the lovely people reading this blog, here's a small excerpt from the book, "The Amazing Results of Positive Thinking" that can reignite your spark when things go down:

Let me know in the comments, what goals, skills and opportunities you are planning to embrace in 2025.

A heartfelt thanks for taking some time to read my blog.
Cheers!

Agent: Content summariser

Saloni Agarwal — Mon, 27 Jan 2025 07:59:26 +0000

This is a submission for the Agent.ai Challenge: Productivity-Pro Agent (See Details)

What I Built

Say hello to your personal summarizer agent that can create a short summary of a webpage content in seconds.
This agent leverages the power of OpenAI o1 model to achieve the above task.All you need to do is paste the Url of the webpage you want a summary of and rest is automated. You can also save the generated summary to a Google Doc for future reference.
So whether you feel too lazy to read something or just want a quick review of a blog/article, this agent has got it all covered.

Demo

agent

Agent.ai Experience

The experience was really amazing. I learned so much about AI agents. Although I am not very happy with this submission since I had really better ideas, I hope I can make something truly useful next time. cheers!

DEV Community: Saloni Agarwal

NanoGem Nail Art Studio 💅🏻

What I Built

Demo

How I Used Google AI Studio

Multimodal Features

Conclusion

Meet Pulse - the ultimate skill coach agent ✨

What I Built

Key features

Demo

Screenshots 📺

How I Used Runner H

Key Steps in the agent workflow

Use Case & Impact

Real World Applications

How it improves existing processes

Social Love 💖

Why I built this

Compiling 2025: learn, build, grow...

Introduction

Main goals 🎯

rasbt / LLMs-from-scratch

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

Build a Large Language Model (From Scratch)

Side Projects ❤️‍🔥

Skills to learn 🖋️ 🎹

New habits 🍀

Conclusion

Agent: Content summariser

What I Built

Demo

Agent.ai Experience